CN101320520B - Vehicle detection method and equipment thereof - Google Patents
Vehicle detection method and equipment thereof Download PDFInfo
- Publication number
- CN101320520B CN101320520B CN200810116944XA CN200810116944A CN101320520B CN 101320520 B CN101320520 B CN 101320520B CN 200810116944X A CN200810116944X A CN 200810116944XA CN 200810116944 A CN200810116944 A CN 200810116944A CN 101320520 B CN101320520 B CN 101320520B
- Authority
- CN
- China
- Prior art keywords
- image
- sift
- characteristic
- vehicle
- invariant features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 22
- 238000006243 chemical reaction Methods 0.000 claims description 35
- 239000013598 vector Substances 0.000 claims description 35
- 238000007621 cluster analysis Methods 0.000 claims description 9
- 238000012706 support-vector machine Methods 0.000 claims description 8
- 238000005286 illumination Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000000052 comparative effect Effects 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 abstract description 6
- 230000009466 transformation Effects 0.000 abstract description 5
- 239000000523 sample Substances 0.000 description 27
- 230000001939 inductive effect Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000004744 fabric Substances 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000004804 winding Methods 0.000 description 2
- FAPWRFPIFSIZLT-UHFFFAOYSA-M Sodium chloride Chemical compound [Na+].[Cl-] FAPWRFPIFSIZLT-UHFFFAOYSA-M 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000012468 concentrated sample Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000008014 freezing Effects 0.000 description 1
- 238000007710 freezing Methods 0.000 description 1
- 230000001771 impaired effect Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000011780 sodium chloride Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a vehicle detection method and a device thereof. The method comprises collecting road surface images/video, extracting the scale-invariant feature transformation (SIFT) characteristic in the road surface images/video, orderly processing the scale-invariant feature transformation (SIFT) characteristic, and transmitting the characteristic to a classifier through the network; wherein, the classifier is prepared through training in advance, and is used for the vehicle detection of the road surface images through the scale-invariant feature transformation (SIFT) characteristic. The vehicle detection method adopts a SIFT operator to describe the characteristic of the vehicle; the detection device is embedded into an ordinary monitoring camera/video camera; and the vehicle detection system has the advantages of low cost, easy realization and effective operation.
Description
Technical field
The present invention relates to a kind of vehicle checking method and equipment thereof.
Background technology
In recent years, the vehicle detection technology is widely used in traffic monitoring, automobile assistant driving system and the automated navigation system, has become an important component part in the intelligent transportation monitoring.Existing vehicle detection technology mainly is to utilize the various inductive coils that are laid under the road surface to detect vehicle.
The toroid winding detecting device is traditional traffic detecting device, is the maximum a kind of checkout equipment of present consumption in the world.Vehicle causes the variation of coil magnetic field through being embedded in the toroid winding under the road surface, and detecting device calculates the traffic parameters such as flow, speed, time occupancy and length of vehicle in view of the above, and is uploaded to central control system, to satisfy the needs of traffic control system.
There is following shortcoming in this system: coil must directly be imbedded the track when installing or safeguarding, the traffic meeting is temporarily hindered like this; The joint-cutting of embedding coil has softened the road surface, makes the road surface impaired easily, and especially in the crossroad that signal controlling is arranged, vehicle launch damages when perhaps braking may be more serious; That inductive coil is subject to is freezing, the influence of subgrade settlement, physical environment such as saline and alkaline; Inductive coil is because self measuring principle limits, and when wagon flow is blocked up, following distance is less than 3m the time, and its accuracy of detection reduces significantly, even can't detect.
And the inductive coil in this system receives electromagnetic interference (EMI) easily, and system is huge, and cost is high, is difficult to widespread use under various complex situations.
Summary of the invention
In order to overcome the deficiency of prior art, the invention provides a kind of vehicle checking method and equipment thereof.
In first aspect, the invention provides a kind of vehicle checking method, comprising: gather pavement image/video; Extract yardstick invariant features conversion (SIFT) characteristic in said pavement image/video; Yardstick invariant features conversion (SIFT) characteristic is carried out ordering handle, and be sent to sorter through network, wherein, said sorter draws through precondition; And said sorter carries out vehicle detection according to yardstick invariant features conversion (SIFT) the characteristic road pavement image of handling through ordering.
In second aspect, the invention provides a kind of vehicle detecting system, comprise sorter and vehicle equipment, said sorter and said equipment interconnect through network.Vehicle equipment comprises: video acquisition module is used to gather pavement image/video; Yardstick invariant features conversion (SIFT) characteristic extracting module is used to extract yardstick invariant features conversion (SIFT) characteristic in said pavement image/video; And yardstick invariant features conversion (SIFT) characteristic ordering module, be used for that yardstick invariant features conversion (SIFT) characteristic of extracting is carried out ordering and handle, with pending vehicle detection.Wherein, said sorter
Utilize SVMsDraw through precondition.
In one embodiment of the invention; Ordering in the first aspect present invention is handled and is comprised cluster analysis; Comprise: judge the proper vector of all yardstick invariant features conversion (SIFT) characteristics on the said pavement image and the similarity of each cluster, and be the set of cluster said character representation.
In another embodiment of the present invention, the similarity in the first aspect present invention is judged the similarity that comprises comparative feature vector sum cluster centre value; Wherein, said cluster centre value obtains through in advance the vehicle image sample being carried out clustering processing.
In yet another embodiment of the present invention, the training step in the first aspect present invention comprises: collect some image pattern and some image patterns that do not comprise vehicle that comprise vehicle; Extract yardstick invariant features conversion (SIFT) characteristic in all images sample; Yardstick invariant features conversion (SIFT) characteristic of extracting is carried out ordering to be handled; And utilize through yardstick invariant features conversion (SIFT) characteristic after the ordering processing and come training classifier.
In another embodiment of the present invention, the ordering in the first aspect present invention is handled and is comprised: cross the vehicle image sample is carried out clustering processing and obtains the cluster centre value; Through the similarity of all yardstick invariant features conversion (SIFT) proper vectors on the computed image sample, characteristics of image is expressed as a C dimensional vector with corresponding each cluster centre value
Wherein, f
iBe the two-valued variable of value 0 or 1, if image comprises a proper vector that belongs to i cluster then f
i=1, otherwise f
i=0; Wherein, C is pre-determined number of clusters.
The present invention describes vehicle characteristics with the SIFT operator, and checkout equipment is embedded monitoring camera/video camera commonly used, has realized a kind of low cost, is prone to realize, and effective vehicle detecting system.
Description of drawings
Below with reference to accompanying drawings specific embodiments of the present invention is explained in more detail, in the accompanying drawings:
Fig. 1 is the block diagram of vehicle detecting system according to an embodiment of the invention.
Embodiment
Fig. 1 is the block diagram of vehicle detecting system according to an embodiment of the invention.
As shown in Figure 1, this system comprises video acquisition module, initialization module, SIFT characteristic extracting module, SIFT characteristic ordering module, sorter.
Video acquisition module is used for video streaming image is taken and obtained to road conditions.
Initialization module is used for video streaming image is carried out initialization process, to obtain high-quality image to be checked.
The SIFT characteristic extracting module is used for extracting from image to be checked the SIFT characteristic of vehicle.
SIFT characteristic ordering module is used for that the SIFT characteristic of vehicle is carried out ordering and handles.
Sorter is used for the SIFT characteristic after the ordering is classified, so that whether existing of vehicle is arranged in the detected image.
Should be understood that; The SIFT operator is a kind of based on metric space; Image local feature for convergent-divergent, rotation even the affined transformation of image all remains unchanged is described operator (full name is Scale InvariantFeature Transformation, the i.e. conversion of yardstick invariant features).
Should also be pointed out that; The system architecture that the present invention proposes is used monitoring camera or the camera acquisition video image that is erected at the roadside; Promptly can video acquisition module, initialization module, SIFT characteristic extracting module, SIFT characteristic ordering module be referred to as embedded processing equipment, and can be when practical application with this embedded processing equipment integration in monitoring camera or video camera.Like this, just can independently accomplish work such as feature extraction and ordering, through network treated characteristic is sent to control center again and carry out subsequent treatment through the camera or the video camera of front end.
The implementation procedure of vehicle detecting system of the present invention comprises training stage and detection-phase; The purpose of training stage is a training sorter recited above; And the purpose of detection-phase is to extract the SIFT characteristic and it is carried out ordering, is sent to said sorter so that carry out vehicle detection through network at last.Below, respectively these two stages are set forth in detail.
Training stage
Collect P the positive sample set of image construction that comprises vehicle, N image that does not comprise vehicle formed the negative sample collection.
Align all concentrated samples of negative sample and carry out initialization process, comprise initialization operations such as yardstick normalization, illumination compensation, histogram equalization, so that extract characteristic.Should be pointed out that this initialization process is optional.
Because the present invention uses SIFT operator (full name is Scale Invariant FeatureTransformation, the i.e. conversion of yardstick invariant features) to describe vehicle characteristics, so need utilize the SIFT characteristic extracting module to extract the SIFT characteristic of all training samples.
For i image training sample, use the SIFT operator to detect all unique points in this training sample.(these SIFT unique points have change of scale unchangeability and rotational invariance to suppose to have extracted n SIFT unique point altogether in this width of cloth image training sample; Target can well be described), will generate one 128 characteristic of tieing up at each characteristic point position place so.Therefore, the characteristic of this image training sample is exactly the set of the proper vector of n 128 dimensions
Explain that unique point generally refers to the Local Extremum of grey scale change in the image, promptly this pixel value is all bigger or all little than the pixel value of the point in certain field around it.In general, described unique point all can comprise significant structural information, under various situation, can both embody.
Adopt the SIFT algorithm to come the method for detected characteristics point to be:
If input picture is I, integer type variable σ (value from 0 to a certain preset integer, such as 0～20) is set, each pixel I (x, y) locate to calculate following formula:
Calculate then:
D(x，y，σ)＝L(x，y，kσ)-L(x，y，σ)
K any non-1 the positive integer of value according to actual needs wherein.
Like this, in the image each pixel can calculate D (x, y, σ), then according to this value search Local Extremum as unique point/key point.
Obtain after the key point, take out this key point small images on every side, carry out some and handle, just obtained the proper vector of one 128 dimension.
Since above the SIFT proper vector that draws be unordered, therefore need carry out ordering and handle the characteristic of extracting, concrete grammar is following.
Preferably, we can carry out the ordering that cluster analysis realizes characteristic to the SIFT proper vector of extracting.At first, by hand from the positive sample of P select K (the representational sample of K＜P).Then, the SIFT proper vector { F that uses the SIFT characteristic extracting module to obtain to above-mentioned K sample
(1), F
(2)..., F
(K)Carry out cluster analysis.Can select multiple clustering method, like hierarchical clustering, mean shift cluster etc.Can obtain C cluster through clustering processing, each cluster i obtains a label l
i∈ 1,2 ..., C}, the central value that calculates each cluster just can obtain all proper vector mean value F (1) in this cluster, F (2) ..., F (C).
Should be pointed out that the cluster centre value that draws in the training stage will be preserved, so that in the detection-phase of back, supply the usefulness of ordering processing.
For each image pattern, calculate the pairing SIFT proper vector of all unique points and the similarity of above-mentioned each cluster centre value on this width of cloth image respectively.Similarity is handled can use multiple method for measuring similarity, such as least mean-square error (MSE), minimum average B configuration absolute difference (MAD), maximum match pixels statistics (MPC) or the like.Calculate through similarity, can obtain the pairing cluster label of unique points all on this image pattern.Like this, the characteristic of each image pattern just can be expressed as a C dimensional vector (f
i)
I=1 C, f wherein
iTwo-valued variable for value 0 or 1.If image pattern comprises a proper vector that belongs to i cluster then f
i=1, otherwise f
i=0.
Below, the ordering processing is provided a detailed explanation.
In general, for pictures different, it is different using the SIFT characteristic extracting module to handle the resulting unique point in back.But,,, extract the unique point that draws so and can have certain rules property such as the above-mentioned positive sample image that is used to train if the image of input belongs to same type objects.In brief, can detect unique point in some part of this type objects exactly.Cite a plain example, the vehicle image as positive sample is detected, most of sample wherein all can detect unique point in positions such as the wheel of vehicle, vehicle windows.Therefore, we can carry out cluster analysis (hypothesis just has only two types: wheel, vehicle window) here to detected these unique points.Through cluster analysis, can obtain this average of two types with this as this characteristic of two types.Should be noted that when actual treatment, not knowing to have what classifications is to have public character, counts C so must roughly estimate a classification, so that carry out cluster analysis.
Accomplish after the cluster analysis, for the sample of each input, we compare the unique point and the cluster centre value of this image, just can know to have comprised how many individual parts (such as wheel, the vehicle window of the above-mentioned example) that has general character on this image.Cite a plain example, make C=2, promptly only consider wheel, two parts of vehicle window.Therefore, for the C dimensional vector { f that representes this image pattern characteristic
i}
I=1 C, comprise wheel on (0,1) expression input sample, do not comprise vehicle window, and comprise vehicle window on (1,0) expression input sample, do not comprise wheel.Like this, just original unordered SIFT characteristic has been accomplished the ordering processing.Should be understood that classification counts C and can get other round valuess beyond 2, can think that characteristics of image divides other more classifications.
After collecting the proper vector after the process ordering is handled in each training sample, can select various classifier design method to come the design category device, such as SVMs, nearest neighbor classifier, artificial neural network or the like.The present invention provides a kind of implementation preferably, promptly utilizes SVMs to train the vehicle classification device.
If given 1 sample that belongs to two types respectively: (F
(1), y
1), (F
(2), y
2) ..., (F
(l), y
l), F wherein
(l)The characteristic of representing the 1st sample, y
lThe classification of representing the 1st sample, y
lValue-1 perhaps+1.
SVMs is sought an optimal classification face, and this optimal classification face not only can said two types separates error-free, and will make two types class interval maximum.Through finding the solution constrained optimization problem, can obtain a following optimal classification face:
Wherein, k () is a kernel function, α
iCan obtain through finding the solution quadratic programming (QP) problem.The symbol of function return value shows the classification of input sample F.
After above-mentioned design, import the sorter of accomplishing into control center, control center is connected through network with said embedded processing equipment, so that in detection-phase subsequently, reality image to be checked is classified.
Detection-phase
At first, road conditions is taken, obtained video streaming image through the video acquisition module in the said embedded device.
Preferably, input picture is carried out initialization process (yardstick normalization, illumination compensation, histogram equalization etc.), to obtain high-quality image to be checked.
For any image I to be detected, use a size as the rectangular search window of w * h in image-region from top to bottom, from left to right, carry out moving successively with the step-length of Δ w and Δ h respectively, promptly carry out window search.Comprise two processing in the described window search process, the processing of promptly extracting the SIFT characteristic and the SIFT characteristic being carried out ordering.
Should be understood that; Since the image to be checked in the relative detection-phase of training image sample in the training stage all much smaller (such as; Positive sample has only vehicle image); So the training stage, generally do not adopt above-mentioned search window to scan, but directly carry out feature extraction and characteristic ordering processing to entire image.
In the moving process of rectangular search window, the SIFT characteristic extracting module uses the SIFT operator to detect all unique points in the rectangular search window.Suppose in entire image, to have extracted altogether n
iIndividual SIFT unique point (these SIFT unique points have change of scale unchangeability and rotational invariance, can well describe target) will generate one 128 characteristic of tieing up in the position of each unique point so.Therefore, this image pattern I
iCharacteristic be exactly n
iThe set of the proper vector of individual 128 dimensions
The same with the training stage and since above the SIFT proper vector that draws be unordered, therefore need be to extracting to such an extent that characteristic is carried out ordering and handled.
Handle in order to carry out that better the SIFT characteristic is carried out ordering, preferably, need carry out cluster analysis the SIFT proper vector of extracting.Can directly be utilized in the cluster centre value that the training stage preserves here.
For each image pattern, calculate the pairing SIFT proper vector of all unique points and the similarity of above-mentioned each cluster centre value on this width of cloth image respectively.Similarity is handled can use multiple method for measuring similarity, such as least mean-square error (MSE), minimum average B configuration absolute difference (MAD), maximum match pixels statistics (MPC) or the like, to obtain the pairing cluster label of unique points all on this image pattern.Like this, the characteristic of each image pattern just can be expressed as a C dimensional vector { f
i}
I=1 C, f wherein
iTwo-valued variable for value 0 or 1.If image pattern comprises a proper vector that belongs to i cluster then f
i=1, otherwise f
i=0.Like this, also drawn the ordering characteristic of each image pattern.
At last, through network, Internet protocol (IP network) preferably is sent to the sorter of control center's (obtaining in the training stage) with the proper vector that obtains.The vehicle classification device is classified to image through the proper vector of importing into, judges whether there is vehicle in this image, then judged result is passed back to embedded processing equipment (camera, video camera) through network, thereby has been accomplished vehicle detection of the present invention.
Obviously, under the prerequisite that does not depart from true spirit of the present invention and scope, the present invention described here can have many variations.Therefore, the change that all it will be apparent to those skilled in the art that all should be included within the scope that these claims contain.The present invention's scope required for protection is only limited described claims.
Claims (15)
1. vehicle checking method comprises:
Gather pavement image/video;
Extract yardstick invariant features conversion (SIFT) characteristic in said pavement image/video;
Yardstick invariant features conversion (SIFT) characteristic is carried out ordering handle, and be sent to sorter through network, wherein, said sorter utilizes SVMs to draw through precondition; And
Said sorter carries out vehicle detection according to yardstick invariant features conversion (SIFT) the characteristic road pavement image of handling through ordering;
Said sorter utilizes the precondition step of SVMs to comprise:
Collect some image pattern and some image patterns that do not comprise vehicle that comprise vehicle;
Extract yardstick invariant features conversion (SIFT) characteristic in all images sample;
Yardstick invariant features conversion (SIFT) characteristic of extracting is carried out ordering to be handled; And
Utilize through yardstick invariant features conversion (SIFT) characteristic after the ordering processing and come training classifier;
Said ordering is handled and is comprised:
Through being carried out clustering processing, the vehicle image sample obtains the cluster centre value;
Through the similarity of all yardstick invariant features conversion (SIFT) proper vectors on the computed image sample, characteristics of image is expressed as a C dimensional vector (f with corresponding each cluster centre value
i)
Wherein, f
iBe the two-valued variable of value 0 or 1, if image comprises a proper vector that belongs to i cluster then f
i=1, otherwise f
i=0;
Wherein, C is pre-determined number of clusters.
2. according to the method for claim 1; Wherein, Said ordering is handled and is comprised and comprise cluster analysis: judge the proper vector of all yardstick invariant features conversion (SIFT) characteristics on the said pavement image and the similarity of each cluster, and be the set of cluster with said character representation.
3. according to the method for claim 2, wherein, similarity is judged the similarity that comprises comparative feature vector sum cluster centre value; Wherein, said cluster centre value obtains through in advance the vehicle image sample being carried out clustering processing.
4. according to the method for claim 3, wherein, said similarity is judged one or more in the following method for measuring similarity of employing: least mean-square error (MSE), minimum average B configuration absolute difference (MAD), maximum match pixels statistics (MPC).
5. according to the method for claim 2, wherein, said clustering processing comprises one or more in the following method: hierarchical clustering, mean shift cluster.
6. according to the method for claim 1, comprise that also the pavement image to collecting carries out following initialization process:
Image is carried out yardstick normalization to be handled; And/or
Image is carried out illumination compensation to be handled; And/or
Image is carried out histogram equalization to be handled.
7. according to the process of claim 1 wherein, said network is Internet protocol (IP) network.
8. a vehicle detecting system comprises sorter and vehicle equipment, and said sorter and said equipment interconnect through network, and wherein said vehicle equipment comprises:
Video acquisition module is used to gather pavement image/video;
Yardstick invariant features conversion (SIFT) characteristic extracting module is used to extract yardstick invariant features conversion (SIFT) characteristic in said pavement image/video; And
Yardstick invariant features conversion (SIFT) characteristic ordering module is used for that yardstick invariant features conversion (SIFT) characteristic of extracting is carried out ordering and handles, with pending vehicle detection;
Wherein, said sorter utilizes SVMs to draw through precondition;
Said sorter utilizes the precondition step of SVMs to comprise:
Collect some image pattern and some image patterns that do not comprise vehicle that comprise vehicle;
Extract yardstick invariant features conversion (SIFT) characteristic in all images sample;
Yardstick invariant features conversion (SIFT) characteristic of extracting is carried out ordering to be handled; And
Utilize through yardstick invariant features conversion (SIFT) characteristic after the ordering processing and come training classifier;
Said yardstick invariant features conversion (SIFT) characteristic ordering module comprises:
Through calculating the similarity of all yardstick invariant features conversion (SIFT) proper vectors on the said pavement image and corresponding each cluster centre value, characteristics of image is expressed as a C dimensional vector (f
i)
Module;
Wherein, f
iBe the two-valued variable of value 0 or 1, if an image comprises the proper vector that belongs to i cluster, then a f
i=1, otherwise f
i=0;
Wherein, C is pre-determined number of clusters;
Wherein, said cluster centre value obtains through in advance image pattern being carried out clustering processing.
9. according to Claim 8 vehicle detecting system, wherein said vehicle equipment also comprises:
Initialization module is used for the vehicle image that collects is carried out initialization process.
10. according to the vehicle detecting system of claim 9, wherein said vehicle equipment also comprises:
Image is carried out the module that yardstick normalization is handled; And/or
Image is carried out the module that illumination compensation is handled; And/or
Image is carried out the module that histogram equalization is handled.
11. system according to Claim 8, wherein, said vehicle equipment is a camera.
12. system according to Claim 8 also comprises exercise equipment, said exercise equipment is used to train said sorter.
13. according to the system of claim 12, said exercise equipment comprises:
Collect some image pattern and some modules that do not comprise the image pattern of vehicle that comprise vehicle;
Extract the module of yardstick invariant features conversion (SIFT) characteristic in all images sample;
Yardstick invariant features conversion (SIFT) characteristic of extracting is carried out the module that ordering is handled; And
The module of utilizing yardstick invariant features conversion (SIFT) characteristic after ordering is handled to come training classifier.
14. the system according to claim 12 also comprises:
Initialization module is used for the said image of collecting is carried out initialization process.
15. according to the system of claim 13, wherein, the said ordering module of said exercise equipment comprises:
Through the said image pattern that comprises vehicle is carried out the module that clustering processing obtains the cluster centre value;
Through the similarity of all yardstick invariant features conversion (SIFT) proper vectors on the computed image sample, characteristics of image is expressed as a C dimensional vector (f with corresponding each cluster centre value
i)
Module,
Wherein, f
iBe the two-valued variable of value 0 or 1, if image comprises a proper vector that belongs to i cluster then f
i=1, otherwise f
i=0;
Wherein, C is pre-determined number of clusters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810116944XA CN101320520B (en) | 2008-07-21 | 2008-07-21 | Vehicle detection method and equipment thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810116944XA CN101320520B (en) | 2008-07-21 | 2008-07-21 | Vehicle detection method and equipment thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101320520A CN101320520A (en) | 2008-12-10 |
| CN101320520B true CN101320520B (en) | 2012-07-04 |
Family
ID=40180547
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810116944XA Active CN101320520B (en) | 2008-07-21 | 2008-07-21 | Vehicle detection method and equipment thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101320520B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102651075B (en) * | 2011-02-28 | 2014-04-16 | 中兴智能交通（无锡）有限公司 | Training method for vehicle identification model, and vehicle identification method and apparatus |
| WO2012125687A2 (en) | 2011-03-14 | 2012-09-20 | The Regents Of The University Of California | Method and system for vehicle classification |
| CN102360434B (en) * | 2011-10-09 | 2013-08-21 | 江苏大学 | Target classification method of vehicle and pedestrian in intelligent traffic monitoring |
| CN102695040B (en) * | 2012-05-03 | 2014-04-16 | 中兴智能交通（无锡）有限公司 | Parallel high definition video vehicle detection method based on GPU |
| CN104966087B (en) * | 2015-06-02 | 2018-06-01 | 北京京东尚科信息技术有限公司 | Image SIFT streak features filtering method and device |
| CN105184286A (en) * | 2015-10-20 | 2015-12-23 | 深圳市华尊科技股份有限公司 | Vehicle detection method and detection device |
| CN105335758A (en) * | 2015-11-03 | 2016-02-17 | 电子科技大学 | Model identification method based on video Fisher vector descriptors |
| CN110349415B (en) * | 2019-06-26 | 2021-08-20 | 江西理工大学 | Driving speed measuring method based on multi-scale transformation |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101196996A (en) * | 2007-12-29 | 2008-06-11 | 北京中星微电子有限公司 | Image detection method and device |
-
2008
- 2008-07-21 CN CN200810116944XA patent/CN101320520B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101196996A (en) * | 2007-12-29 | 2008-06-11 | 北京中星微电子有限公司 | Image detection method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101320520A (en) | 2008-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101320520B (en) | Vehicle detection method and equipment thereof | |
| Mithun et al. | Detection and classification of vehicles from video using multiple time-spatial images | |
| CN106683119B (en) | Moving vehicle detection method based on aerial video image | |
| Lai et al. | Lane detection by orientation and length discrimination | |
| CN101916383B (en) | Vehicle detecting, tracking and identifying system based on multi-camera | |
| CN103971097B (en) | Vehicle license plate recognition method and system based on multiscale stroke models | |
| Luvizon et al. | Vehicle speed estimation by license plate detection and tracking | |
| CN106682586A (en) | Method for real-time lane line detection based on vision under complex lighting conditions | |
| Neubert et al. | Beyond holistic descriptors, keypoints, and fixed patches: Multiscale superpixel grids for place recognition in changing environments | |
| CN104951775A (en) | Video technology based secure and smart recognition method for railway crossing protection zone | |
| CN110619279A (en) | Road traffic sign instance segmentation method based on tracking | |
| CN110164139B (en) | System and method for detecting and identifying side parking | |
| Xiang et al. | Lightweight fully convolutional network for license plate detection | |
| CN103353941B (en) | Natural marker registration method based on viewpoint classification | |
| CN114913498A (en) | Parallel multi-scale feature aggregation lane line detection method based on key point estimation | |
| Zhang | Vehicle target detection methods based on color fusion deformable part model | |
| Appiah et al. | Object detection in adverse weather condition for autonomous vehicles | |
| CN109977862A (en) | A kind of recognition methods of parking stall limiter | |
| Xuan et al. | Robust lane-mark extraction for autonomous driving under complex real conditions | |
| Cheng et al. | A novel approach for license plate slant correction, character segmentation and Chinese character recognition | |
| Deepan et al. | Road recognition from remote sensing imagery using machine learning | |
| CN115620149B (en) | Road detection method based on remote sensing image | |
| Yu et al. | An improved phase correlation method for stop detection of autonomous driving | |
| CN114663839A (en) | Method and system for re-identifying blocked pedestrians | |
| Ghosh | A modified YOLO model for on-road vehicle detection in varying weather conditions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210223 Address after: No. 602, 6th floor, shining building, 35 Xueyuan Road, Haidian District, Beijing 100083 Patentee after: BEIJING ZHONGXINGTIANSHI TECHNOLOGY Co.,Ltd. Address before: 100083, Haidian District, Xueyuan Road, Beijing No. 35, Nanjing Ning building, 16 floor Patentee before: Vimicro Corp. |
|
| TR01 | Transfer of patent right |